Summary of Deepfake Detection Without Deepfakes: Generalization Via Synthetic Frequency Patterns Injection, by Davide Alessandro Coccomini et al.
Deepfake Detection without Deepfakes: Generalization via Synthetic Frequency Patterns Injection
by Davide Alessandro Coccomini, Roberto Caldelli, Claudio Gennaro, Giuseppe Fiameni, Giuseppe Amato, Fabrizio Falchi
First submitted to arxiv on: 20 Mar 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Deepfake detectors are typically trained on large sets of pristine and generated images, which limits their ability to generalize; they excel at identifying deepfakes created through methods encountered during training but struggle with those generated by unknown techniques. Our proposed learning approach enhances the generalization capabilities of deepfake detectors by exploiting the unique “fingerprints” that image generation processes consistently introduce into the frequency domain. We train detectors using only pristine images, injecting crafted frequency patterns that simulate the effects of various deepfake generation techniques without being specific to any. We evaluated our approach using diverse architectures across 25 different generation methods. The models trained with our approach demonstrated state-of-the-art deepfake detection and superior generalization capabilities compared to previous methods. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Deepfake detectors are like super smart eyes that can spot fake pictures. But they get confused when they see new, weird ways of making fake pictures. We want to make them better at recognizing all kinds of fake pictures. To do this, we came up with a new way to train these detectors using only normal pictures and adding special patterns to them that mimic how fake pictures are made. We tested our idea on many different types of picture-making techniques and it worked amazingly well! The detectors got really good at spotting fake pictures no matter how they were made. |
Keywords
* Artificial intelligence * Generalization * Image generation
